The emergence and rapid development of generative AI such as ChatGPT, presents new opportunities and challenges for primary school English listening and speaking learning. It is of great significance to study the roles and collaborative activities among teachers, students, and machines in primary school English listening and speaking learning, to effectively improve students’ English listening and speaking ability. This study pointed out the problems in primary school English listening and speaking learning such as vague learning objectives, unclear human-machine collaboration relationship, and lack of student-centered activity design. This study used the methods of literature research, theoretical deduction, and quasi-experiment. It clarified the elements of primary school students’ English listening and speaking ability and developed a measurement scale. Based on the theories of second language acquisition and human-machine collaborative learning, this study concluded the general process of primary school English listening and speaking learning, explained the human-machine collaboration mechanism, then conducted a framework for designing human-machine collaborative learning activities to enhance primary school students’ listening and speaking ability. Using the design framework, 40 primary school students from GZ City underwent English listening and speaking learning. The results showed that their listening and speaking ability and overall class performance had significantly improved. The findings from this study may provide instructional design guidance and practical references for enhancing primary school English listening and speaking learning.